Capital & Industrial Strategy
Top Line
Anthropic's revenue surge and multibillion-dollar CoreWeave infrastructure deal signal intensifying competition with OpenAI, though the companies report revenue differently through cloud partnerships, complicating direct comparison.
Japan commits an additional $4 billion to Rapidus chipmaking, bringing total support past $16 billion in a high-stakes bet to break into AI semiconductor supply chains dominated by TSMC, which just posted 35% revenue growth from sustained AI chip demand.
Power infrastructure emerges as the binding constraint on AI deployment, with PJM seeking 15 gigawatts of emergency capacity while Big Tech backs next-generation nuclear and xAI faces legal opposition over its Mississippi power plant.
Amazon signals strategic pivot away from Nvidia chips toward proprietary alternatives, marking direct competition between cloud hyperscalers and the GPU supplier that currently dominates AI training infrastructure.
Key Developments
Anthropic Secures Infrastructure and Closes Revenue Gap with OpenAI
Anthropic signed a multibillion-dollar agreement with CoreWeave to rent data center capacity for Claude, the companies announced, with CoreWeave CEO Michael Intrator confirming the deal's scale though exact financial terms remain undisclosed. CoreWeave stock rose 11% on the news, which followed a $21 billion commitment from Meta one day earlier. The infrastructure agreement comes as Anthropic closes in on OpenAI's US business revenue, driven by strong enterprise adoption of Claude Code products, though Semafor reports the companies report revenue differently when selling through cloud partners like AWS, Google, and Microsoft, making direct comparison difficult.
The infrastructure expansion addresses what Anthropic characterized as increasing demand for its AI services, according to Bloomberg. The timing is notable given Anthropic temporarily banned OpenClaw's creator from accessing Claude after pricing changes last week, suggesting the company is actively managing capacity constraints and usage patterns as it scales.
Japan Escalates Chipmaking Industrial Strategy with $16 Billion Rapidus Bet
Japan approved ¥631.5 billion ($4 billion) in additional subsidies for Rapidus Corp., bringing total government support past $16 billion in what Bloomberg characterizes as a signature project widely regarded as a long shot to break into the intensely competitive AI chipmaking arena. The subsidy acceleration comes as incumbent TSMC posted a 35% revenue jump to a new record high, benefiting from sustained demand for advanced semiconductors from customers including Apple and Nvidia.
The Rapidus funding represents a national industrial strategy to establish domestic advanced semiconductor capacity in AI chips, a sector where Japan has no current meaningful position. The scale of public capital committed underscores government assessment that market forces alone will not produce the supply chain diversification deemed strategically necessary, particularly as AI chip demand continues to drive TSMC revenue growth.
Power Infrastructure Emerges as Critical Bottleneck for AI Deployment
PJM Interconnection LLC is seeking 15 gigawatts of new power supplies in an emergency proposal to address potential electricity shortages stemming from the AI boom, while Big Tech companies are putting financial heft behind next-generation nuclear power as demand surges, according to Reuters. The power constraint is already generating political friction, with Elon Musk's xAI facing fresh legal opposition from environmental groups after landing a permit for a massive power plant in Mississippi to support its operations, now owned by SpaceX.
The scale of PJM's emergency procurement request — 15 gigawatts represents roughly 15% of current PJM capacity — indicates grid operators view power availability as a near-term constraint on data center expansion rather than a planning consideration for the medium term. The divergence between rapid AI deployment timelines and multi-year power infrastructure development cycles creates strategic risk for companies unable to secure dedicated generation capacity.
Cloud Providers Signal Strategic Shift Away from Nvidia Dominance
Amazon took a jab at Nvidia, signaling that a shift away from its chips has begun as cloud hyperscalers develop proprietary alternatives. Semafor characterizes this as a signal that Nvidia is in direct competition with the frontier AI labs it already supplies. The strategic tension comes as Amazon, Google, and Microsoft all invest in custom silicon to reduce dependence on Nvidia's GPUs while simultaneously remaining major Nvidia customers for near-term capacity.
The public messaging from Amazon represents a notable shift from the cloud providers' previous positioning that custom chips would complement rather than substitute for Nvidia hardware. The competitive dynamic is complicated by the fact that cloud providers both compete with and supply infrastructure to frontier AI labs, creating multiple layers of strategic conflict as the market matures.
Signals & Trends
Revenue Recognition Complexity Obscures True Competitive Position in Enterprise AI
The difficulty in comparing Anthropic and OpenAI revenue, stemming from different approaches to reporting sales through cloud partnerships with AWS, Google, and Microsoft, suggests the enterprise AI market lacks standardized metrics for assessing competitive position. This opacity benefits incumbents by making it harder for investors and customers to evaluate true market dynamics, while creating risk for strategics conducting M&A due diligence. As enterprise procurement increasingly flows through existing cloud relationships rather than direct contracts, the ability to assess which AI provider is winning may remain opaque until cloud platforms disclose partner revenue in greater detail.
AI Powerhouses Threaten Data Processing Incumbents Through Vertical Expansion
Semafor reports that software companies' shares have taken a hit as AI models move in on customer data processing, a signal that foundation model providers are expanding vertically into application-layer functions previously performed by specialized software. This pattern suggests AI infrastructure providers view owning more of the value chain as strategically preferable to enabling a robust ecosystem of third-party applications, reversing the platform dynamics that characterized previous technology shifts. The market reaction indicates investors believe incumbent data processing firms lack defensible moats against foundation model providers who can integrate similar functionality directly.
Infrastructure Partnerships Replace Equity Stakes as Preferred Strategic Relationship Structure
The CoreWeave-Anthropic and CoreWeave-Meta deals, totaling over $21 billion in committed spending, represent a shift toward multi-year infrastructure commitments rather than equity investments or acquisitions as the preferred mechanism for securing compute capacity and customer relationships. This structure allows AI companies to secure capacity without diluting ownership while enabling infrastructure providers to finance buildout against contracted revenue. The model contrasts with earlier AI investment patterns dominated by equity stakes from cloud hyperscalers in frontier labs, suggesting the market is maturing toward more conventional customer-supplier relationships with long-term contracts replacing strategic ownership as the primary form of partnership.
Explore Other Categories
Read detailed analysis in other strategic domains